Haze control: Reticle/environment interactions at 193nm
05/01/2004
Until recently, the sole measure of a reticle's cleanliness prior to use was inspection. As new mask processes are introduced and exposure wavelengths shrink, the science of cleaning has taken a giant leap forward. Residuals on the mask surface and in the environment around the mask become more relevant: reactions in DUV radiation can be fatal to the reticle's performance. Maskmakers must now be proactive and find ways to verify the robustness of new materials and carefully assess process changes. An investigation focusing on haze formation and accelerated lifetime testing is presented and the relationship between surface and environmental reactions is examined.
In preparing for the ramp-up of 193nm products, DPI sought to identify potential roadblocks by surveying the industry. In tandem, an internal laboratory was set up with 193nm (ArF) radiation to study product reliability. There was already speculation and skepticism within the industry regarding pellicle and compact material compatibility on prototype 193nm masks.
End users were asked to allow DPI to conduct routine monitoring of reticles at different exposure levels. The results were alarming. At <1kJ of total energy exposed, a significant growth was found encompassing the glass side of the mask (Fig. 1). Less commonly, contamination was also seen along the outer edges on the patterned side, but safely outside of the pellicle. The haze was insignificant to the process so far, but concern arose that a seemingly subdued growth left unchecked might ultimately affect mask transmission properties or eventually extend into the pattern itself. Additionally, the backside growth was invisible to typical patterned side through-pellicle inspections used in the field.
Contamination analysis
An outsourced lab using Raman spectroscopy and time-of-flight secondary-ion mass spectrometry (TOF-SIMS) performed haze analysis. Multiple test runs produced peaks at 975cm-1 pointing to ammonium sulfate ((NH4)2SO4) as the leading constituent. This was referenced against an unexposed control mask. Cyanuric acid was not detected, contrary to studies in which it was correlated to haze growth at 193nm within the patterned area under the pellicle [1]. The analysis also ruled out another potential key player — organic airborne molecular contamination — which is also believed to contribute to hazing through condensation and adsorption [2].
The formation of salts at shorter wavelengths due to available anions and cations has been well documented within the IC industry. Ammonium (NH4+) and sulfates (SO4-2), the reaction components of the discovered salt, are commonly used in mask-cleaning chemistries and are prevalent within the environment.
Mask environment
To assist in a laboratory reproduction of the growth, an analysis was first undertaken to measure the mask environment within a production wafer-fab lithography tool (see table). This included a sampling of inorganic materials (including acids and bases) and high boiling-point organics. Air samples purged through ultrapure water were set out for a four-hour period and then processed using ion chromatography. A concurrent sampler purged air through a porous collection medium and was analyzed using a gas chromatograph equipped with a mass selective detector and thermal desorption system. Samples were measured against control blanks that used the same collection medium and were handled in the same way, but with no volume drawn.
 |
Overall, the low concentrations of acids and bases were impressive. Nitrous acid, a weak acid, was markedly higher. Primary players, such as ammonia, were kept at sub-1ppb levels within the tool and to exceptionally low levels even in the ambient.
Experimental
Optical variables adjusted in the lab — including laser fluence and frequency, temperature, and pressure — were matched to fab lithography-tool specifications and testing commenced using a similar environment to that measured. Variables were manipulated so that testing could be accelerated while providing an accurate response, compared to masks assessed in the field. This was verified through quantitative and qualitative analyses of mask surfaces in addition to tool and visual inspections. Haze produced in the lab varied only by grain size, yielding a slightly coarser contamination. These adjustments dramatically accelerated feedback times to just days vs. the weeks or months mandated by steppers in the field.
A detailed array of mask manufacturing parameters was assembled next for testing to determine possible links. This was derived from a commonality study that cross-examined process and clean chemistries, mask-blank materials, optical platform, resist, pellicle type, and compact materials. Exposure environment was a coupled variable controlled at the lab.
Results from the matrix of experiments validated the following:
- Environmental photochemical reaction at 193nm. Compounds in the environment near the exposed mask combine and migrate to the surface.
- Surface photochemical reaction at 193nm. Residual chemistry left on the surface after a clean combine on the surface.
The results were independent and nullified other factors included (such as pellicle and compact materials) as tested against the total energy used in this study.
 Figure 2. Reproduced haze signatures of a) the environment, b) the environment plus residuals, and c) only residuals. |
In follow-up experiments, the environmental and surface effects were separated for illustrative purposes. Using a newly developed mask clean, purge baffles within the exposure chamber were directed over only the edges of the mask to localize airborne contamination during exposure (Fig. 2a). In another example: a typical "production-clean" mask was exposed in a controlled environment with less than detectable levels of sulfates and amines (Fig. 2c). The combined environmental and mask residual conditions leave overlapping trademarks on the same mask (Fig 2b). This experiment highlighted the independence of each as a contributing mechanism.
Controlling haze
Surface residuals from mask cleans play a significant role in 193nm haze growth. As a manufacturer of 193nm photomasks, DPI is interested in the consistency and durability of its reticles. Programs were initiated to address this issue on two fronts: minimizing residuals through advanced cleaning processes, or alternatively, a post-process virgin capping layer applied to the backside of the mask.
Finding the lower limits residual "window" was accomplished through repeated experimental cleans and exposure cycles. Understanding the complete neutralization of chemistries and how surface compounds coexisted post-clean was key to reducing residuals to barely traceable levels on both the patterned and glass sides of the mask. DPI has applied this methodology to improve clean reliability in spin-spray and bath technologies alike.
 Figure 3. Schematic representation of the residual levels required for 248nm and 193nm haze-free reticles exposed in a clean environment. |
New target residual levels required on 193nm masks were found to be a magnitude <248nm mask allowable levels (Fig. 3). Standardized measurement comparisons were made using a mask fully immersed in ultrapure water and processed using ion chromatography. Aggressive and repeated radiation testing has confirmed that these levels will remain benign throughout the lifetime of the mask.
Capping the backside of the mask is a patent-pending approach to eliminating surface residuals by replacing the surface altogether. This process was successfully demonstrated using a fluoropolymer film applied to the backside of the mask as a final step before exposure. The thickness was tuned to maximize transmission and the material properties provided additional optical benefits.
Looking ahead
As 193nm volume is expected to dramatically increase, the integration of new processes for these products has already begun. In an effort to further optimize mask longevity in the wafer fab, we are continuing our investigation by studying the environmental effects and durability requirements of mating components such as the pellicle, compact (packaging), and the wafer-fab lithography tool, which can also contribute to contamination growth. This is being accomplished in part due to a recently upgraded exposure test bed allowing for extremely accurate control and feedback of environmental conditions.
Conclusion
Every wavelength and material change introduces new challenges to product reliability, where degradation modes occur from time and use. In this study, we were able to separate and recreate two prevalent components, environmental conditions and mask residuals, which contribute to growth contamination on 193nm photomasks. The fundamentals presented in this study have taken on clarity from improved analytical methods and change control practices, such as in situ lifetime testing.
Product reliability testing is used throughout DPI and will soon be extended to the newly opened Advanced Mask Technology Center (AMTC) as the industry approaches the 65nm technology node. It is a key element needed to qualify new mask processes, materials, and point-of-use changes before reaching mainstream production in wafer fabs. Thanks to determination and an open relationship between supplier and customer, 193nm haze was properly dismantled and a new cleaning standard was established for the upcoming generation of photomasks.
References
- K. Bhattacharyya, W. Volk, B. Grenon, D. Brown, J. Ayala, "Investigation of Reticle Defect Formation at DUV Lithography," BACUS 2002.
- A.J. Dallas, D. Arends, K. Fischer, J. Joriman, K. Graham, et al., "Protecting the DUV Process and Optimizing Optical Transmission," Metrology, Inspection, and Process Control for Microlithography XIV, Vol. 3998, 2000.
For more information, contact Eric V. Johnstone, a backend R&D engineer at DuPont Photomasks Inc., 400 Texas Ave., Round Rock, TX 78664; ph 512/310-6285.